PREDICTING NOX EMISSIONS FROM BIOMASS COFIRING

Stephen Niksaa, Guisu Liua, Larry Felixb, P. Vann Bushb, and

Douglas M. Boylanc aNiksa Energy Associates, 1745 Terrace Drive, Belmont, CA 94002,

neasteve@pacbell.net bSouthern Research Institute, P. O. Box 55305, Birmingham, AL,

35255-5305 cSouthern Company Services, Inc., P. O. Box 2641, Birmingham,

AL, 35291-8195 Introduction

In theory, biomass cofiring should reduce NOX emissions simply because most forms of biomass contain less fuel-N than the coals they displace. The yields of volatiles are generally much higher from biomass than from coal, and volatiles contain many compounds that can effectively reduce NO in fuel-rich flame zones. Biomass also decomposes faster than coal, which tends to steepen near-burner temperature profiles and enhance the ignition characteristics. These factors are expected to improve NOX emissions by more than the displacement of coal-N. But the database from pilot- and full-scale testing exhibits mixed results, including substantial representation for no effect, for less reduction than the displacement of fuel-N, and for enhanced reduction over the displacement of fuel-N.

The premise for our work is that uncontrolled variables are responsible for the ambiguities in the existing database. A testing campaign was staged to characterize the most likely parameters, including the biomass form, coal type, biomass injection configuration, burner staging, and furnace stoichiometry. The database from the testing campaign was then interpreted with simulations based on full chemistry for fuel decomposition, volatiles combustion including fuel-N conversion, soot conversion, and char burnout. Once the model predictions were demonstrated to be accurate, the simulation results were interrogated to determine how the biomass affected NOX emissions. We find that two factors determine whether biomass cofiring will reduce NOX emissions: (1) Does the abundance of gaseous volatiles, not soot, from biomass reduce away the NO formed near the burner; and, (2) Is significantly less char-N released into downstream flame zones by the addition of biomass. This paper emphasizes the interpretation of the database with detailed chemical reaction mechanisms. More thorough descriptions of the simulation methodology are available1,2. Testing Program All tests were conducted in Southern Research Institute’s (SoRI) Combustion Research Facility (CRF). The CRF consists of fuel handling and feeding systems, a vertical refractory lined furnace with a single up-fired burner, a horizontal convective pass, heat exchangers and conventional exhaust cleaning devices. Emissions from this system have been qualified against those from full-scale, tangential-fired furnaces for the operating conditions imposed in this work3. The 8.5 m by 1.07 m (i.d.) cylindrical furnace handles gas velocities from 3 to 6 m/s, and residence times from 1.3 to 2.5 s. The nominal firing rate is 1.0 MWt, which was fixed for all cases in this work. A single burner generates a core of pulverized fuel and primary air surrounded by weakly swirled secondary air. For all cases in this paper, the biomass was co-milled with the coal and the mixture was directly fed into the burner. Overfire air (OFA) was injected through 4 off-radius ports located 4.6 m down the furnace. All datasets include cases with 0 (unstaged) and 15 % OFA. In the unstaged series, the OFA was combined with secondary air. Air

feedrates were varied to produce exhaust (wet) O2 levels between 2.5 and 5 %.

Table 1. Fuel Properties

SD SG JR GL PR JW

Proximate, as rec’d

Moisture 9.5 15.2 19.3 5.8 1.9 0.8

Ash 0.4 29.5 5.4 6.6 15.1 14.6

Volatiles 78.1 47.6 39.6 33.7 33.2 20.0

Carbon 11.9 7.7 35.7 54.0 49.9 64.6

Ultimate, daf wt. %

C 49.8 56.1 74.9 81.8 83.4 89.5

H 6.1 5.4 4.9 5.0 5.5 4.6

O 43.9 35.7 18.8 10.0 7.5 3.3

N 0.2 2.4 0.9 2.0 1.8 1.7

S 0.0 0.4 0.4 1.1 1.8 0.9

<dP>, µm 163 173 29 53 48 34

Fuel quality was varied by firing four diverse coals with two

forms of biomass. Nominal properties of the six primary fuels appear in Table 1 in order of increasing rank, from left to right. Moisture levels are highest for the low-rank fuels, especially switchgrass (SG) and subbituminous coal (JR). Ash levels are widely variable and especially high for biomass SG and for the hv bituminous (PR) and low volatile bituminous (JW) coals. Whereas it appears that the volatility of sawdust (SD) is much higher than SG’s, on a dry-ash-free (daf) basis, their volatiles yields are almost identical. The daf-volatiles contents of the coals fall by more than a factor of two over this suite of samples, which will definitely affect the conversion of coal-N into NOX. Carbon contents increase and oxygen contents decrease for fuels of progressively higher rank. The pair of biomass samples represents most of the range of elemental compositions seen for diverse forms of biomass. The represented range of coal rank, from subbituminous through lv bituminous, is similarly broad. The hydrogen, nitrogen, and sulfur levels are not rank-dependent. Whereas almost all biomass contains little nitrogen and sulfur, sample SG contained the most nitrogen of any of the fuels, due to its decomposition before firing. The particle size distributions (PSDs) are typical utility grinds for the coals, except for the much finer grind of JR, whereas the biomass grinds are much coarser, as expected.

Computer Simulations

It is not currently possible to incorporate detailed chemical reaction mechanisms into conventional CFD simulations of pulverized fuel (p. f.) flames. Since chemistry in the gas phase, especially volatile-N conversion chemistry, was suspected to play a dominant role in NOX production during biomass cofiring, Niksa Energy Associates developed a new computational approach for this application based on our “ChemNet Post-Processing (CNPP)” method. The CNPP method first generates an equivalent network of idealized reactor elements from a conventional CFD simulation. The reactor network is a computational environment that accommodates realistic chemical reaction mechanisms; indeed, mechanisms with a few thousand elementary chemical reactions can now be simulated on ordinary personal computers, provided that the flow structures are

Prepr. Pap.-Am. Chem. Soc., Div. Fuel Chem. 2003, 48(2), 643

Equivalent Reactor Network restricted to the limiting cases of plug flow or perfectly stirred tanks. The network is “equivalent” to the CFD flowfield in so far as it represents the bulk flow patterns in the flow. Such equivalence is actually implemented in terms of the following set of operating conditions: The residence time distributions (RTDs) in the major flow structures are the same in the CFD flowfield and in the section of the reactor network that represents the flow region under consideration. Mean gas temperature histories and the effective ambient temperature for radiant heat transfer are also the same. The entrainment rates of surrounding fluid into a particular flow region are evaluated directly from the CFD simulation. To the extent that the RTD, thermal history, and entrainment rates are similar in the CFD flowfield and reactor network, the chemical kinetics evaluated in the network represents the chemistry in the CFD flowfield. Whereas this paper emphasizes the application to biomass cofiring, separate publications explain the simulation methodology in greater detail1,2,4.

As seen in Fig. 1, the bulk flow patterns are delineated in the regions assigned by CNPP from the CFD simulation. The baseline PR flame consists of a core, mixing layer, ERZ (which is inconsequential), OFA-region and burnout (BO) region. All other flames have the same regions. The CSTR network from the CNPP analysis of the baseline PR flame also appears in Fig. 1. The networks for all other CRF flames have similar branches and feedstreams but appreciably different quantitative specifications. In the network, the flame core has been subdivided into two regions. The devolatilization zone covers the upstream portion of the core in which volatiles are being released from the fuel suspension and burned with primary air. Since the primary stream is extremely reducing, no residual O2 leaves the devolatilization zone. The NOX reduction zone covers the downstream portion of the core in which only the N-species are converted under the influence of water gas shifting, due to the absence of O2. The CSTR-series for the mixing layer and the OFA zone represent the mixing of secondary and tertiary air streams, respectively. But there are no additional flows into the CSTR-series for the BO zone.

CFD Simulations

CFD simulations of the CRF were performed by Reaction Engineering International (REI) for all tests with PR coal and various biomass. Cases with JR and JW were performed at SoRI, using REI’s Configurable Fireside Simulator (CFS) for the CRF. The CFS imposes a fixed computational grid on the calculations, and is therefore suitable for parametric case studies with the same firing configurations. No CFD were available for cases with the second hv bituminous coal (GL) because these flame structures were expected to be very similar to the PR flames.

The RTD for this particular core was deconvoluted into one component for 16 CSTRs-in-series and another for plug flow with respective mean residence times of 138 and 193 ms. The plug flow component represents the near-axial fluid motion under the influence of particle drag, and the CSTR-component represents flow with significant radial velocities. The networks for other fuels usually have only one flow channel for the entire core, and the bulk flow pattern is plug flow. The RTD for the mixing layer was matched with a series of 19 CSTRs, and that for the OFA zone was represented by 6 CSTRs-in-series. The BO zone is essentially in plug flow.

The most surprising feature in the CFD simulations is that all fuel particles remained on the furnace axis throughout this furnace. Neither mixing in the near-burner zone nor radial OFA injection disperses the particles off their original trajectories. Near the burner, the primary air stream was significantly expanded by the release of volatiles from the fuel suspension and by thermal expansion. This expansion zone delineates a fuel-rich core from the outer, annular flow of secondary air. Nominal residence times in the core range from 120 to 170 ms. The expansion of the primary flow promotes entrainment of secondary air into the core, because some of the secondary flow penetrates the expansion boundary. In addition, a portion of secondary air is entrained into the core as soon as it passes the edge of its delivery tube in the fuel injector. Together, these entrainment mechanisms almost instantaneously mix about 20 % of the secondary air into the primary flow.

Note that entrainment into the various CSTR-series is represented as a series of discrete additions over several reactors in the series. Volatiles are entrained into the series for the first part of the flame core; secondary air is entrained into the series mixing layer; and tertiary air is entrained into the series for the OFA zone. The addition rates of volatiles were specified from the stand-alone devolatilization simulation with the thermal histories of particles from the CFD simulation. The specific addition rates of the air streams were specified from the continuous entrainment profiles evaluated from the CFD simulation. Reaction Mechanisms

The mixing layer between the core and secondary air streams gradually expands until it contacts the furnace wall midway to the OFA ports. An external recirculation zone (ERZ) forms in the corner bounded by the outer boundary of the secondary air stream, but is too weak to entrain particles or appreciable amounts of air or fuel compounds. As the fuel compounds in the flame core contact the secondary air stream, they mix and burn in an expanding mixing layer. This layer completely surrounds the core near the burner inlet, and fills the entire furnace downstream of the core. The most distinctive feature of the mixing layer is that the temperature profile across the layer in the normal direction passes through a maximum value which is essentially the same around the entire circumference of the core. Maximum gas temperatures approach 1700°C in cases where fuels of the highest heating values were fired without OFA, and 1600°C in cases with 15 % OFA. Residence times in the mixing layer to the OFA location vary from 500 to 600 ms. The four air jets from the OFA ports do not penetrate onto the centerline. They also do not fill the entire flow cross section. Downstream of the OFA ports, the flow relaxes to a plug flow pattern that carries ash and exhaust into an exhaust system. The total residence time from the burner to the furnace exit is approximately 2.5 s.

The devolatilization submodel, called FLASHCHAIN, determines the complete distribution of volatile compounds from almost any p. f., and also predicts the yield and elemental composition of char5. When combined with a swelling factor correlation and a correlation for the initial carbon density in char, it specifies all the necessary char properties for a char oxidation simulation. Hence, the complete distribution of volatiles, including gaseous fuels and soot, and all char properties are completely determined from only the fuel’s proximate and ultimate analyses.

The reaction mechanism for chemistry in the gas phase contains 444 elementary reactions among 66 species, including all relevant radicals and N-species6. All rate parameters were assigned independently, so there are also no adjustable parameters in the submodel for gas phase chemistry. The soot chemistry submodel depicts several important effects. As soot burns, it directly competes for the available O2 and also consumes O-atoms and OH that would otherwise sustain homogeneous chemistry. Soot also promotes recombinations of H-atoms and OH that could also sustain NO homogeneous chemistry7. And soot reduces NO directly into N2.

Prepr. Pap.-Am. Chem. Soc., Div. Fuel Chem. 2003, 48(2), 644

Mixing Layer OFA Zone Burnout ZoneCore

ERZ

…

Char Primary

Air

BURNOUT ZONE 1460 ms Exhaust Gases

LOI + Flyash

DEVOLATILIZATION ZONE, 65 ms

NO REDUCTION ZONE 8 CSTRs, 73 ms

… …

Volatiles

DEVOLATILIZATION ZONE8 CSTRs, 65 ms

Secondary Air

…

MIXING LAYER 19 CSTRs, 508 ms

OFA ZONE 6 CSTRs, 156 ms

Char Primary

Air

Tertiary Air (OFA)

NO REDUCTION ZONE, 128 ms

Figure 1. (Top) CNPP regions delineated from the CFD simulation for the baseline PR flame, and (Bottom) equivalent reactor network.

Char burning rates are determined by thermal annealing, ash encapsulation (of low-rank chars), and a transition to chemical kinetic control. The Char Burnout Kinetics (CBK) Model includes all these effects, and depicts the impact of variation in gas temperature, O2 level, and char particle size within useful quantitative tolerances8. However, it is not yet possible to specify the initial char reactivity within useful tolerances from the standard coal properties. We must calibrate this value with LOI predictions or some other suitable index on combustion efficiency. The submodel for char-N conversion is subject to a similar calibration requirement (with NO emissions), compounded by its simplistic mechanistic premise; viz., that a fixed fraction of char-N is converted into NO at the overall burning rate throughout all stages of char oxidation.

Calibration and Extrapolation Procedures The fraction of char-N converted to NO during char oxidation

was assigned to fit the NOX emissions from the coal-only baseline flames for each coal sample. These same values were applied in all cofiring simulations. Values for biomass were unnecessary because biomass chars contain no nitrogen.

Only 13 CFD simulations were developed for the CRF under this project, yet almost 300 tests were simulated with detailed reaction mechanisms. All but one of the cases with CFD simulations had 15 % OFA, and were for the coal-only baselines or the high loading of biomass. Extrapolations to operating conditions without CFD were based on perturbations to the flame temperature profiles and air entrainment rates. All these procedures were established for the coal-only baseline flames to describe the reported impact of furnace stoichiometry and staging level on the NOX emissions. Once established, the same extrapolation procedures were then applied without adjustment to the cofiring cases.

To summarize, the initial char reactivity and the fraction of char-N converted to NO can only be specified from calibration procedures, whereby these parameters are adjusted to match the predicted LOI and NOX emissions to reported values for a single set of operating conditions. Then the same values should be imposed for all other operating conditions. Except for these two parameters, all other model parameters can be assigned from the fuel’s proximate and ultimate analyses within useful quantitative tolerances, or directly adopted from literature.

Results

The discussion in this section moves through the mechanistic basis for NOX reduction via biomass cofiring, beginning with the distinctive devolatilization behavior of the various fuels in the testingprogram. Then the NOX predictions are evaluated in comparisons with data, followed by our interpretations for the major trends based on the detailed flame structure of CRF flames.

Prepr. Pap.-Am. Chem. Soc., Div. Fuel Chem. 2003, 48(2), 645

Table 2. Distributions of Secondary Pyrolysis Products and Char Properties.

SD SG JR GL PR JW Volatiles, daf wt. % Wt. Loss 86.1 86.0 65.2 56.5 59.8 39.7 Soot 4.3 13.4 30.1 33.7 37.9 26.8 CH4 7.1 7.4 0.7 0.4 0.5 0.3 C2H2 2.2 1.3 1.5 1.0 1.3 2.3 C2H4 1.4 1.5 0.0 0.0 0.0 0.0 H2 2.1 1.7 3.4 3.6 4.0 3.5 CO 48.4 41.5 12.9 7.2 6.1 1.7 CO2 8.2 8.0 6.4 2.2 1.7 1.0 H2O 12.1 7.7 7.6 4.9 4.3 1.8 HCN 0.0 0.0 1.26 2.47 2.24 1.52 NH3 0.24 2.90 0.0 0.0 0.0 0.0 H2S 0.0 0.44 0.42 1.17 1.91 0.96 Char Comp., daf wt. % C 94.7 97.1 98.9 98.4 98.5 98.2 H 3.4 2.5 0.5 0.5 0.4 0.5 O 1.9 0.4 0.0 0.0 0.0 0.0 N 0.0 0.0 0.4 1.1 1.0 1.26 S 0.0 0.0 0.1 0.1 0.1 0.0 Char ash, wt.% 2.5 75.7 15.9 14.2 30.9 21.8 Char size, µm 97.6 103.6 29.9 59.1 54.1 41.7

Table 3. Evaluation of Predicted NOX for 3.5 % O2 with 15 % OFA.

Series Fuel NOX, ppm @ 3 % dry O2 Pred. Mes’d 1 PR hv bit 325 328 PR/10%SD 255 271 PR/20%SD 236 245 PR/15%SG 277 269 PR/20%SG 258 260 5 GL hv bit 372 360 GL/10%SD 365 273 GL/20%SD 318 327 6 GL hv bit 368 356 GL/10%SG 399 340 GL/20%SG 349 336 7 JR subbit 263 240 JR/10%SG 203 185 JR/20%SG 177 142 JR/10%SD 221 191 JR/20%SD 212 189

12 JW lv bit 419 422 JW/5%SD 416 420 JW/10%SD 417 389 JW/20%SD 436 364

13 JW lv bit 429 450 JW/5%SG 422 455 JW/10%SG 442 440 JW/20%SG 404 407

Predicted Devolatilization Behavior. Since the heating rate of

primary air in the CRF is very fast, the primary devolatilization products are instantaneously converted into secondary volatiles pyrolysis products. The predominant transformation during secondary pyrolysis is the conversion of tar into soot, with simultaneous release of tar-O as CO, tar-H as H2, and most of the tar-N as HCN. In addition, all aliphatic hydrocarbons are converted into CH4 and C2H2, which can add to the soot phase during the latest stages. The predicted distributions of secondary pyrolysis products are collected in Table 2. Total volatiles yields are the same for both forms of biomass and, at 86 daf wt. %, much higher than the yields from any of the coals. The biomass product distributions are dominated by CO, with substantial amounts of hydrocarbons, especially CH4, and CO2 and H2O. H2 is another major fuel compound from biomass. But there is surprisingly little soot, considering that tar, the soot precursor, is 25 to 45 % of the daf fuel mass released during primary devolatilization. The reason is the abundance of tar-O, which approaches 40 % of the tar mass. This oxygen converts most of the tar into CO rather than soot during secondary pyrolysis. Essentially all the fuel-N is released as NH3 during secondary pyrolysis. The abundance of NH3 with SG, and its higher soot yield and lower CO yield, are the major difference between the two biomass forms.

significant variation in char properties is among the char-ash levels. The values for SG and PR are definitely high enough to inhibit char oxidation during the latest stages of burnout, according to the ash inhibition mechanism in CBK. The mean char sizes are disparate for the biomass and coal chars, but more similar than the whole coal values because biomass shrinks and the coals swell during devolatilization, especially the bituminous coals

Evaluation of Predicted NOX Emissions. The reported impact of fuel quality on NOX emissions is extremely complex. It depends on which coal is fired, which biomass is cofired, which cofiring level is imposed, and which injection configuration is used. Table 3 presents the measured and predicted NOX emissions for all cases with co-milled fuel injection and 15 % OFA that had 3.5 % exhaust O2. Cofiring with 10 % SD always reduced NOX, except with JR coal (and with GL coal with 0 % OFA). With the same injection, cofiring with 20 % SD reduced NOX with JR and GL but not with JW. With PR/20 % SD, NOX was reduced with 15 % OFA (but not with 0 % OFA). Cofiring with 20 % SG was effective on PR and JW, but notwith GL and JR. With 10 % SG, cofiring was effective with JR, but not with GL and JW; PR was not tested at this condition. In contrast, the secondary pyrolysis products from all the coals

are dominated by soot which is, by far, the most abundant product. The total hydrocarbon yields are comparable from all coals, but less than a fourth of the hydrocarbon yields from the biomass. Hydrogen yields are also comparable, and double those from biomass. The yields of the oxygenated gases diminish with coals of progressively higher rank, in accord with the trend in the coal-O levels. But even the highest CO yield from JR coal is only about one-quarter the CO yield from the biomass. The only N-species is HCN although, in actuality, a minor amount of NH3 may have been released from JR coal (but none of the others). The N-species yields are directly proportional to the coal-N levels, as expected.

As seen in Table 3, the predicted emissions for PR hv bit coal (Ser. 1) are within experimental uncertainty for all biomass cofiring combinations. They correctly depict the more-than-20 % reduction in NOX with the highest loadings of both biomass forms, and a proportional decrease in NOX for progressively higher biomass loadings. No discrepancy exceeds 16 ppm. Similarly, the predictions for JR subbit. (Ser. 7) depict the proportional reductions in NOX with higher biomass loadings, and the greater effectiveness of SD compared to SG. In this series no discrepancy exceeds 35 ppm. Cases with the JW lv bit. coal are within experimental uncertainty for SG co-firing (Ser. 13), which did not reduce NOX at either of the lower switchgrass loadings. The predicted NOX reduction with 20 % SG (Ser. 12) was under 10 %, in accord with the measured value. But with SD cofiring on JW coal (Ser. 12), the predictions show no significant NOX reduction for any SD loading, at

Char compositions are very similar among all six fuels, except that the most abundant heteroatoms in biomass chars are H and O. versus H and N in the coal-derived chars. The char-N levels are negligible in biomass chars. They are comparable but lower for the coal-derived chars and certainly not negligible. Perhaps the most

Prepr. Pap.-Am. Chem. Soc., Div. Fuel Chem. 2003, 48(2), 646

odds with the reported reductions up to 11 % for loadings of 10 and 20 %.

Figure 2. Predicted (Curves) and measured (data points) NOX emissions for 3 staging levels and a range of exhaust O2 levels for PR/15 % SG flames.

Predicted NOX emissions with GL coal (Ser. 5 and 6) are not as

accurate as all others, and the reported behavior is more complex. With SD cofiring, the reported extent of NOX reduction is significantly greater for a 10 % loading than for 20 %. With SG cofiring, the same extent of NOX reduction was reported for both loadings. But the predicted NOX emissions are actually higher than the coal-only baselines with both biomass forms. Unfortunately, no CFD simulations were available for any of the GL test cases, so the specifications for PR-coal were applied in all GL-cases. A breakdown in this extrapolation procedure may be responsible for the discrepancies in the predictions, although they may also reflect a flaw in the mechanisms for this particular chemical environment.

The accuracy of the extrapolation procedures for furnace stoichiometry and staging level are apparent in Fig. 2. This set of predictions covers the full ranges of furnace stoichiometry (which determines exhaust O2 level) and degree of air staging in the test program. The predictions are based on the procedures developed to fit the NOX emissions for the coal-only baseline tests over the same domain of conditions, starting with only a single CFD simulation for 3.5 % O2 and 15 % OFA. The same procedures were then applied without modification to the biomass cofiring cases, such as the 15 % SG on PR in Fig. 2. The predictions are within experimental uncertainty except, perhaps, for the highest O2 level with 15 % OFA. Notwithstanding, it is evident that these extrapolations did not introduce intolerable uncertainties into the predictions, and CNPP can be successfully applied with many fewer CFD simulations than test conditions. Discussion

The predicted structure of the flame core for the PR/20%SD flame appears in Fig. 3. In contrast to SG, SD has almost no fuel-N. This cofired flame still generates 25% less NOX than the PR-only baseline flame, which is slightly more than the reduction expected for the removal of fuel-N alone. The time scale in Fig. 3 depicts the core as well as the first quarter of the mixing layer. For this particular test, devolatilization is completed within 80 ms, and the flow leaves the core at 118 ms. The total residence time in the mixing layer is 476 ms, but only the first 150 ms are shown in Fig. 3.

The S. R. value for the gas phase falls sharply while volatiles are released into the flow, making it more reducing. It then relaxes to an ultimate value 0.864, which is significantly more reducing than both the PR-only baseline and the PR/20%SG flame. Although SD and SG have identical volatiles yields, the gas phase in the SD-core becomes more reducing because more CO and less soot are produced by the primary volatiles from SD.

2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5150

200

250

300

350

400

450

500

550

30 % OFA

15 % OFA

0 % OFA

Co-Milled Injection, Ser. 185 % Pratt/15 % Switchgrass

NO

X, pp

m @

3%

dry

O2

Exhaust O2, wet

Initially, the CO concentration surges during the ignition period, then increases more gradually during the oxidation of char and soot. Its ultimate value and the persistence of H2 reflect water gas shifting once all O2 has been consumed. The maximum CO concentration is double that in the PR/20%SG flame. The H2 mass fraction persists at roughly 1000-2000 ppmw across the entire core. Moreover, hydrocarbons, especially C2H2 (not shown), persist at 1000 ppmw or more across the entire devolatilization zone. This is the only flame core with an appreciable amount of hydrocarbons in the presence of NO, although their concentration is still much lower than those of CO and H2.

As for the other flames, almost half the char burns out in the core, but hardly any soot burns out in this particular core. The more reducing character of the gas phase in this core imparts several distinctive features to the N-species conversion chemistry. The N-speciation is dominated by HCN and NO, as for the PR-only baseline flame. The NH3 released by the SD is rapidly converted to HCN and NO within 40 ms. But NO does not accumulate at the expense of HCN, as in both of the other flame cores. Instead, the HCN concentration surges while NO accumulates. Some prompt N-fixation mechanisms involving N2 in air must be responsible because the total maximum amount of fixed-N species is double the maximum value in the PR/20%SG flame, and the SG-cofired flame has significantly more volatile-N. NO reduction begins at 75 ms but by the exit of the core, there is still 1390 ppmw HCN and 465 ppmw NO, which are both much higher than in either of the other flames. Early in the mixing layer, NO reduction accelerates while the HCN concentration plummets. The NH3 concentration reaches 92 ppm before vanishing with the HCN concentration. The ultimate NO concentration after both other fixed-N species have been eliminated is only 143 ppmw, which is 25 ppm lower than the analogous level in the PR/20%SG flame. Even though there was a higher concentration of fixed-N species in the core, the greater reducing potential yielded a lower NO concentration in the mixing layer, after chemistry in the gas phase was exhausted. Since the extents of char burnout at this point are comparable for the PR/20%SG and PR/20%SD flames, the 20 ppmw reduction for the PR/20%SD flame persists in the exhaust emissions.

A surge in the extent of soot burnout coincides with the entrainment of secondary air. Due to the high maximum temperatures in the mixing layer, the extent of soot oxidation eventually overtakes the extent of char oxidation at 220 ms. The soot burns out in the mixing layer, whereas char is carried over into the OFA and BO regions.

The structures of the coal-only and the biomass cofired flames in the CRF are qualitatively similar: All exhibit a very rapid surge in the NO level immediately after injection, due to the ignition of volatiles under the very lean conditions associated with the primary streams during the initial stages of devolatilization. But as more volatiles are released, the gas phase becomes progressively more reducing, which enables the early NO to be reduced into HCN and NH3. The extent of reduction is determined by the S. R. value among the gaseous species only and the residence time available before the core fluid is exposed to secondary air in the mixing layer. All fixed- N species are rapidly converted into NO and N2 at the beginning of the mixing layer, which then sustains the oxidation of soot and char.

Prepr. Pap.-Am. Chem. Soc., Div. Fuel Chem. 2003, 48(2), 647

0

200

400

600

800

1000

1200

1400

0

5

10

15

20

25

S.R.

TGAS

PR/20%SD CR & ML15 % OFA, 3.5% O2

Gas

Tem

pera

ture

, C

S.R

.

0 50 100 150 200 250 3000

5

10

15

20

0

1

2

3

H2

CO

O2

O2,

mas

s%

Residence Time, s

PR/20%SD CR & ML15 % OFA, 3.5% O2

CO

& H

2,mas

s%

0

200

400

600

800

1000

1200

1400

1600

0

25

50

75

100

125

150

NH3NO

HCN

NO

& H

CN

, ppm

w

PR/20%SD CR & ML15 % OFA, 3.5% O2

NH

3, pp

mw

0 50 100 150 200 250 3000

20

40

60

80

0.0

0.2

0.4

0.6

0.8

Ent. Fr.

XSOOT

XCHAR

PR/20%SD CR & ML15 % OFA, 3.5% O2

X CH

AR &

XSO

OT,

%

Residence Time, s

Entra

inm

ent F

r.

Figure 3. Structure of the core of the PR/20%SD flame showing, in counterclockwise order from the upper left, the operating conditions, major species, char and soot burnout, and N-species.

From this point onward, chemistry in the gas phase is inconsequential, and the conversion of some of the char-N into NO supplements the NOX inventory.

Two factors are primarily responsible for the significant NOX reduction observed for CRF flames with biomass cofiring. First, there is more volatile matter from biomass and it contains a much smaller contribution from soot. Consequently, the near-burner S. R. values for the gas phase are significantly richer than in coal-only flames. Under richer conditions, near-burner NO will be reduced away and a greater proportion of the fixed-N species will be converted into N2. CRF flames provide sufficient residence times in the flame cores for NO reduction and fixed-N conversion to N2. Second, a significantly lower percentage of the total fuel-N will remain in the char beyond the near-burner zone, where its partial conversion to NO is inevitable. If the biomass contains nitrogen, it releases all of it in the near-burner zone. If the biomass contains no nitrogen, then the inventory of char-N beyond the flame core is reduced in proportion to the biomass loading. In either case, there is less char-N beyond the point where NO can be reduced by chemistry in the gas phase.

The first factor is especially sensitive to fuel quality. Indeed, even the NOX emissions from the four coal-only flames are ordered according to the volatiles yields and contributions from soot. While the volatiles yields decrease and the soot contributions increase for coals JR, PR, GL, and JW (cf. Table 2), the baseline NOX emissions in the staged flames increased from 240 to 338 to 358 to 436 ppm (cf. Table 3). Biomass cofiring mitigated these differences by enhancing volatiles yields and by reducing the soot contributions. The net effect was significant NOX reduction for the cofired flames,

even when the biomass supplements the inventory of fuel-N. The extent of reduction was directly proportional to the biomass loading with both forms of biomass on JR and PR coals. But it was only proportional to the SG loading in flames cofired with JW coal; cofiring JW with sawdust was ineffective at all but the highest loadings. Cofiring GL coal reduced NOX, but the magnitude was disproportionate at 10 % loadings with both biomass forms.

Acknowledgement. This work was partially sponsored by an award (DE-FC26-00NT40895) under the Biomass Cofiring Opportunities Program administered by the National Energy Technology Laboratory. References (1) Niksa, S.; and Liu G.-S., Fuel, 2002, 81(18),2371-85. (2) Niksa, S.; and Liu, G.-S., “Advanced CFD Post-Processing for P. F.

Flame Structure and Emissions,” 28th Int. Technical Conf. on Coal Utilization and Fuel Systems, 2003, Coal Technology Assoc., Clearwater, Fl, March.

(3) Monroe, L. S. et al., 1995, EPRI/EPA 1995 Joint Symp. On Stationary Combustion NOX Control, Book 3, EPRI, Palo Alto, CA.

(4) Niksa, S.; and Liu, G.-S. Proc. Comb. Inst., 2002, 29, to appear. (5) Niksa, S. Combust. Flame, 1995, 100, 384-394. (6) Glarborg, P.; Alzueta, M. U.; Dam-Johansen, K.; and Miller, J. A.

Combust. Flame, 1998, 115, 1-27. (7) Pedersen, L. S.; Glarborg, P.; et al. Combust. Sci. Technol., 1998, 132,

251-314. (8) Hurt, R. H.; Sun J.-K.; and Lunden, M. Combust. Flame, 1997,

113(1/2), 181.

Prepr. Pap.-Am. Chem. Soc., Div. Fuel Chem. 2003, 48(2), 648